Review of Industrial Organization 20: 115–126, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.

115

Market Power and/or Efficiency: A Structural Approach RIGOBERTO A. LOPEZ
, AZZEDDINE M. AZZAM and CARMEN LIRÓN-ESPAÑA Department of Agricultural and Resource Economics, University of Connecticut, Storrs, CT 06269-4021, U.S.A. E-mail: [email protected]

Abstract. This article separates oligopoly-power and cost-efficiency effects of changes in industrial concentration and assesses their impact on output prices in 32 food-processing industries. Empirical results indicate that although concentration induces cost efficiency in one-third of the industries, oligopoly-power effects either dominate cost efficiency or reinforce inefficiency, resulting in higher output prices in most industries. The article also provides fresh econometric estimates of oligopoly power and economies of size for the industries in question. Key words: Economies of scale, food processing, industrial concentration, industrial organization, oligopoly power. JEL Classifications: L00, L11, L13, L66.

In his review of the new empirical industrial organization (NEIO) literature, Bresnahan (1989) concludes that although NEIO models are useful in measuring market power, they are not as useful in guiding policy when market structure rather than conduct is the policy target. First, he argues that the narrow focus of NEIO studies on single and often heavily concentrated industries is too limited to be useful in mapping structure into conduct and performance. Cross-industry studies, on the other hand, provide information over a wider range of industries and, despite their well-known problems, continue to influence policy (Salinger, 1990). Second, Bresnahan also argues that since market power is imputed rather than observed in NEIO models, its relationship to observable structural and behavioral variables in often unclear to policy makers. Thus, to make NEIO findings more policy-relevant, studies should consider a wider range of industries and incorporate observable structural measures of interest to policy makers, such as industrial concentration.
The authors are grateful to two anonymous referees as well as seminar participants at the Eco-

nomics Research Service (USDA), Universit´e Laval and University of Basel. Financial support was provided by the USDA CRREES special grant No. 00-34178-9036, the University of Connecticut and the University of Nebraska. This is Scientific Contribution No. 1951 of the Storrs Agricultural Experiment Station.

116

RIGOBERTO A. LOPEZ ET AL.

This article develops an NEIO model that formally incorporates measures of industrial concentration and separates out the oligopoly-power from the costefficiency effects of concentration on output prices.1 The model is the oligopoly analogue of Azzam’s (1997) oligopsony model, which extends Appelbaum’s (1982) model to formally include industrial concentration. The separation of the two effects is not only of academic interest but is also of public policy concern because the basic problem facing antitrust authorities is that of a tradeoff between efficiency and market power (Williamson, 1968; Perry, 1984; Bian and McFetridge, 2000). That is, whether or not concentration is in the public interest depends critically on whether or not the cost-efficiency gains through concentration offset the welfare losses from greater market power. The model is applied to 4-digit SIC data on 32 U.S. manufacturing industries over the 1972–1992 period. The econometric results provide fresh estimates of oligopoly power and economies of size in these industries and reveal that Cournot behavior is widespread. The empirical findings also indicate that although costefficiency effects from concentration are important in one-third of the industries, in nearly every case the oligopoly-power effects dominate or reinforce cost inefficiencies, resulting in higher output prices. The few exceptions where concentration is beneficial to buyers are in the fat-and-oil sector which is characterized by strong economies of size and product homogeneity. Finally, this analysis shows that NEIO models that bridge the gap between conduct and market structure can be useful for policy making decisions, especially those targeting industries based on efficiency vs. the consumer’s interest. I. The Model The starting point is an industry of N firms producing a homogeneous good Q requiring factors xr for r = 1, . . . , k and facing a derived market demand curve Q = f (p, z),

(1)

where p is output price and z is a vector of demand shifters. Profit maximization by the j th firm yields the supply relation p=−

∂Cj (qj , w, t) sj (1 + φj ) + , η ∂qj

(2)

where sj = qj /Q is the j th firm’s market
share, η = (dQ/dP )(1/Q) is the semielasticity of demand (η < 0), φj = d ni=j qi /dqj is the j th firm’s conjectural 1 A large number of studies have tested the relationship between efficiency proxies and price-

cost margins (e.g., Demsetz, 1973; Martin, 1988; Rosenbaum, 1994) while others have tested more explicitly the relationship between costs and concentration (Peltzman, 1977; Dickson, 1994) or the ad hoc relationship between price and market structure (e.g., Cotterill, 1986). A few recent studies have separated oligopsony power from cost efficiency using structural models (Azzam, 1997; Azzam and Schroeter, 1995) but studies of this type focusing on oligopoly power are lacking.

117

MARKET POWER AND/OR EFFICIENCY

variation, Cj (.) is the cost function, w is a vector of factor prices, and t is the state of technology. By Shephard’s Lemma, the derived demand for the rth factor by the j th firm is given by xrj =

∂Cj (qj , w, t) ∂wj

for r = 1, 2, . . . , k.

(3)

Following Olson and Shieh (1989), and Baffes and Vasavada (1989), the cost function is assumed to take the modified generalized Leontief form    1/2 1/2 αij wi wj + qj t γi wi + qj2 βi wi , (4) Cj (q, w) = qj i

j

i

i

where αij , γI , and βI are parameters. Multiplying through Equations (2) and (3) by sj , using (4), and summing across the industry yields, respectively, the industrywide analogue of the supply relation p=−

H (1 + η

)

+

 i

1/2

1/2

αij wi wj

+t

j

 i

γi wi + 2H Q



βi wi , (5)

i

and factor demand  Xr = αij Q i j



wj wi

1/2 + tγi + H Qβi

for r = 1, 2, . . . , k,

(6)


where H = j sj2 is
the Herfindahl index, is the industry (weighted) conjectural variation and Xr = j xrj is total industry employment of the rth factor. By virtue of the expression for the semi-elasticity of demand, the demand function (1) takes the semi-logarithmic form  δi zi , (7) ln Q = δ0 + ηp + i

where η, δ0 and δi are parameters. The first term on the right-hand side of the supply relation in (5) is the mark-up over marginal cost. Its magnitude depends on the level of concentration (H ), the semi-elasticity of demand (η), and the type of market conduct ( ). If conduct is competitive, then = −1 and the markup is zero. For Cournot, = 0 and the markup is −H /η. For conduct that is less competitive than Cournot, 0 < ≤ (1/H ) − 1 and the upper bound on the markup is −1/η. However, for noncompetitive conduct, concentration affects both the markup and marginal cost. Note that the commonly-used conjectural variation elasticity (Appelbaum, 1982) can be defined as ∗ = (1 + )H which ranges between 0 and 1, the price elasticity of demand is given by η∗ = ηP , and the industry oligopoly power is defined by L = − ∗ /η∗ .

118

RIGOBERTO A. LOPEZ ET AL.

Following Azzam (1997), is treated as constant.2 Thus, differentiating (5) with respect to H yields the decomposed effects of concentration on output price: dp (1 + =− dH η

)

+ 2Q



βi wi ,

(8)

i

where the first term on the right-hand side is the oligopoly-power effect and the second is the cost-efficiency effect (or the effect of a rise in concentration on marginal cost). A measure for the cost elasticity with respect to output is given by the ratio of industry marginal cost to average cost ecy =

A + 2H QB , A + H QB

(9)





1/2 1/2 + t i γi wi and B = where A = i j αij wi wj i βi wi . Note that ecy depicts economies of size and is the inverse of the degree of returns to scale. If B = 0, constant returns exist, and the only effect of rising concentration on price is through oligopoly power. If B > 0, diseconomies of scale exist, and a rise in concentration raises prices through a rise in both oligopoly power and costs. When economies of scale are present (B < 0), the effect of a rise in concentration on price can be positive, negative, or zero, depending on whether the oligopoly-power effect is larger than, smaller than, or the same as the cost-efficiency effect.3 II. Empirical Procedures The model is operationalized with data at the 4-digit SIC (1987 definitions) level for the 1972–1992 period. The econometric model is based on Equations (5), (6), and (7) depicting pricing behavior, input demand equations, and the output demand equation. Although (5) is the main equation of interest, the input and output demand equations impose stronger theoretical restrictions and assist in identifying the corresponding parameters in the pricing equations. The model assumes three 2 Following the work of Stigler (1964) and Stalhammar (1991), two additional specifications of

were tested. One was to let = θ0 + θ1 ln H in order to allow industry conduct to vary with concentration. Another test was the inclusion of imports and exports. Although these extensions provided some interesting insights in some cases, they deteriorated the results of interest. Given the focus and scope of this article, was therefore treated as a constant. 3 Note that these effects of concentration are for a constant level of output and that higher concentration leads to lower, higher, or no change in costs if there are increasing, decreasing, or constant returns to scale, respectively. By fixing output, a rise in the Herfindahl index implies a change in the distribution in output across firms, with more output being produced by the larger firms leading to lower industry cost in the presence of economies of scale. Note that the econometric model measures economies of scale on a market share-weighted industry cost function and that technical change (t) is assumed to affect the industry marginal cost intercept, not the slope.

MARKET POWER AND/OR EFFICIENCY

119

variable inputs: Materials (XK ), labor (XL ), and capital (XK ).4 Thus, the estimating model consists of five equations: the pricing equation, three input demand equations, and the output demand equation. The latter is assumed to be a function of output price (P ), income (y) and a trend variable (t), where price and income are deflated by the consumer price index (d). The endogenous variables are Q, P , XK , XL , and XK . The exogeneous variables are WK , WL , WK , y, t (= 1 for 1972 through 21 for 1992), d, and H . The parameters to be estimated are , η, the αij s, the βj s, the .i s, d0 , and the /0 and δi s.5 The main data source for prices and quantities of outputs and inputs was the online National Bureau of Economic Research database of Bartelsman et al. (2000) on U.S. manufacturing industries. Following the U.S. Department of Labor (1997, p. 107), we define the rental price of capital in dollars per unit of real capital stock. Due to lack of data on the price of capital at the 4-digit SIC level, all industries are assumed to face the same rental prices but each to have different levels of capital stock. Therefore, the rental price of capital was computed by dividing the cost of capital services (provided electronically by the Bureau of Labor Statistics) divided by total capital assets at the 2-digit SIC level. Due to data limitations, we use an instrumental variable for H for the years when it was not available (see Appendix for details). Given the endogeneity of output quantity and the price of output, the system of five equations is estimated with non-linear 3SLS using the SHAZAM 7.0 software. The results for 32 food industries are presented below. III. Empirical Results 1. C ONDUCT, D EMAND , AND E CONOMIES OF S IZE Table I presents selected estimated parameters: , η, L and 0cy . The null hypothesis for conduct is = −1 (competitive behavior) and for the elasticity of cost with respect to output 0cy = 1 (constant returns to scale). An alternative conduct hypothesis tested is Cournot behavior ( = 0), given its common use in empirical analysis. As usual, one, two and three asterisks next to the estimated coefficients indicate significance at the one, five and 10 percent levels, respectively. Note that the indicated statistical significance for an 0cy in Table I are relative to −1 and 1 (rather than the usual null hypothesis of 0), respectively. 4 See Bartelsman and Gray (1996) for precise definitions of these variables. Following the U.S. Department of Labor (1997), capital services are assumed to be proportional to capital stocks. The derivation of the price of capital is discussed below. The material inputs include raw materials (agricultural, packaging, and certain services) as well as energy. The latter accounted for less than 2% of variable costs in all cases. 5 The structural model contained 17 coefficients, which we estimated with 21 observations. Since the sample is small and the standard errors for nonlinear models are only approximately correct for small samples, the statistical significance of the coefficients should be interpreted with caution.

120

RIGOBERTO A. LOPEZ ET AL.

Table I. Selected parameter estimates, U.S. food processing industries, 1972–1992. SIC

Industry

2011 2013 2015 2022 2023 2024

Meat packing Saus. & prep. meats Poultry & egg proc. Cheese Dry/cond. & ev. milk Ice cream & fruit desserts Fluid milk Canned specialties Canned fruit & veg. Dried fruit & veg. Pickles, sauces, etc. Cereal breakfast foods Rice milling Prep. flour & doughs Wet corn milling Prep. feeds Cane sugar Cane sugar refining Beet sugar Candy & conf. prods. Chocolate & cocoa Cottonseed oil mill Soybean old mill Vegetable oil mill An./mar. fats & oils Malt beverages Wines & brandy Extracts & syrups Roasted coffee Manuf. ice Macaroni & spaghetti Food preparations

2026 2032 2033 2034 2035 2043 2044 2045 2046 2048 2061 2062 2063 2064 2066 2074 2075 2076 2077 2082 2084 2087 2095 2097 2098 2099

s.e.

η

s.e.

L

s.e.

0¯cy

s.e.

−0.603∗∗∗a 0.726 −0.482 −0.865 0.099

0.147 0.808 0.329 0.129 0.412

−0.175∗∗∗ −0.243∗∗ −0.135 −0.095 −0.387∗∗∗

0.059 0.112 0.083 0.090 0.173

0.099∗∗∗ 0.144∗∗∗ 0.106∗∗∗ 0.094∗∗∗ 0.197∗∗∗

0.011 0.017 0.017 0.015 0.059

0.950∗∗∗b 0.998 0.489 1.011 1.037

0.013 0.015 0.011 0.011 0.039

−0.872 3.154∗∗ −0.785∗∗∗ −0.397 −0.871∗∗∗ 0.063 −0.863∗∗∗ −0.495∗∗∗ −0.829∗∗∗ −0.918∗∗∗ −0.422∗ −0.752∗∗∗ −0.996∗∗∗ −0.751∗∗∗ 0.512 −0.989∗∗∗ −0.255 0.993∗∗∗ −0.636∗∗∗ −0.002 −0.682∗∗∗ −0.016 −0.557∗∗∗ −0.853∗∗∗ 6.515∗∗∗

0.634 1.528 0.073 0.509 0.196 0.687 0.037 0.145 0.153 0.039 0.227 0.121 0.011 0.098 0.342 0.018 0.287 0.013 0.125 0.315 0.116 0.727 0.141 0.098 2.255

−0.049 −0.444∗∗∗ −0.404∗∗∗ −0.168 −0.114 −0.433 −0.259∗∗∗ −0.191∗∗∗ −0.314∗ −0.196∗∗∗ −0.096∗∗∗ −0.121∗∗∗ −0.026 −0.191∗∗∗ −0.465∗∗∗ −0.052 −0.332∗∗∗ −0.023 −0.221∗∗∗ −0.095∗∗∗ −0.864∗∗∗ −0.433 −40.349∗∗∗ −0.165∗∗∗ −0.876∗∗∗

0.241 0.157 0.101 0.139 0.171 0.270 0.052 0.031 0.188 0.045 0.035 0.028 0.040 0.040 0.077 0.012 0.121 0.034 0.063 0.029 0.186 0.317 0.105 0.025 0.218

0.097∗∗∗ 0.200∗∗∗ 0.125∗∗∗ 0.118∗∗ 0.086∗∗∗ 0.815∗∗∗ 0.182∗∗ 0.252 0.063 0.090∗∗ 0.087∗∗∗ 0.178∗∗ 0.041 0.215∗∗∗ 0.272∗∗∗ 0.058 0.201 0.035 0.198∗∗∗ 0.417∗∗∗ 0.090∗∗∗ 0.204∗∗∗ 0.286∗∗∗ 0.147 0.241∗∗∗

0.023 0.017 0.026 0.020 0.023 0.046 0.033 0.056 0.041 0.037 0.018 0.071 0.056 0.064 0.043 0.036 0.039 0.024 0.042 0.049 0.027 0.088 0.030 0.095 0.035

1.063∗∗ 0.916 1.214∗∗∗ 1.104∗∗∗ 1.195∗∗∗ 0.884∗∗∗ 1.553∗∗∗ 0.256∗∗ 1.310∗∗∗ 1.055 0.991 0.918 1.013 0.879∗ 1.071 1.294∗∗∗ 0.839∗∗ 0.997 0.841∗∗∗ 0.732∗∗∗ 1.103∗∗∗ 0.971 1.529∗∗∗ 1.077 0.908∗∗∗

0.026 0.019 0.032 0.023 0.029 0.022 0.060 0.062 0.062 0.036 0.019 0.070 0.063 0.074 0.052 0.046 0.027 0.022 0.040 0.042 0.029 0.025 0.058 0.118 0.032

0.103 0.503∗∗∗ 0.042 0.799∗∗∗ 0.081 0.329∗∗∗ 0.043 0.918∗

0.040 0.045

0.272 2.453∗∗∗

0.367 −0.354∗∗∗ 0.921 −0.363∗∗∗

Levels of statistical significance are represented by ∗ (10%), ∗∗ (5%) and ∗∗∗ (1%). The standard errors (s.e.) are indicated next to the estimated coefficients. The null hypothesis for is −1 (perfect competition), while the null hypothesis for 0cy is 1 (CRS). These results are based on the joint estimation of Equations (5), (6) and (7).

MARKET POWER AND/OR EFFICIENCY

121

The results in Table I indicate that the estimated s ranged from −0.996 for SIC 2062 (very close to the lower limit of −1) to 6.515 for SIC 2097 (well below its maximum limit of 63 given by H −1 − 1). Twenty of 32 industries (63%) had non-competitive industry conduct at the 5 percent level. It should be noted that the estimated s for 20 of the industries also rejected the null hypothesis of Cournot behavior ( = 0) at the 5% level. The derived conjectural variation elasticities ranged from 0.00068 for SIC 2062 to 0.2445 for 2075. The estimated oligopoly Lerner indexes (evaluated at mean values of P ) ranged from 0.035 in SIC 2075 to 0.815 in SIC 2035. Twenty-seven of the 32 industries (84statistically different from perfect competition at the 5% level. The number of industries with non-competitive mark-ups is much larger than those indicating non-collusive conduct from only estimated s. As shown by Nevo (2000) in the breakfast cereal industry, it is possible to have non-collusive behavior and yet have a strong degree of oligopoly power and level of mark-up. As a case in point, the soybean oil industry (SIC 2075) was the most collusive, based on the conjectural variation elasticities; however, this industry had the lowest mark-up as indicated by the Lerner index, suggesting that in some cases concentration is a way to survive when industries are operating on small margins. Nonetheless, the results are consistent with a myriad of studies that have found that most food processing industries are non-competitive.6 In terms of economies of size (calculated at mean values of data), the results reveal that approximately 12 industries (38%) have significant economies of size, 9 (28%) have significant diseconomies of size, and the remaining 11 (34%) did not reject the constant returns to size hypothesis at the 5% level. The economies of size parameters, which are critical for the evaluation of cost efficiency effects, appear to be reasonable and consistent with previous findings.7

2. E STIMATED M ARKET P OWER AND E FFICIENCY E FFECTS Table II presents the results based on Equation (8) for the separate effects of changes in the Herfindahl index on oligopoly-power, cost efficiency, and output price. These effects were calculated and tested at mean values of the data. In terms of oligopoly power, at the 10% level, the results indicate that 26 of the 32 industries (81%) significantly increase oligopoly-power as concentration 6 See, for example, Bhuyan and Lopez (1997). In particular, the estimated Lerner index for the meatpacking industry (average = 0.099) is between the ones estimated by Schroeter (1988) for beef (0.04) and by Schroeter and Azzam (1990) for beef and pork (0.14). Likewise, the mark-ups estimated by Morrison-Paul (2000) correspond to Lerner indexes between 0.05 and 0.20 between 1971 and 1991. 7 The average estimate of Morrison-Paul (2000) for long run economies of size in meat packing is 0.965 (cf. 0.95 in Table I). The average degree of economies of size in Table I is 0.997 which is nearly CRS, somewhat above the 0.834 estimated by Bhuyan and Lopez (1997).

122

RIGOBERTO A. LOPEZ ET AL.

Table II. Impacts of industrial concentration on oligopoly power, cost and output price.

SIC

Industry

Impact of H on Oligopoly power s.e.

2011 2013 2015 2022 2023 2024 2026 2032 2033 2034 2035 2043 2044 2045 2046 2048 2052 2061 2062 2064 2064 2066 2074 2075 2076 2077 2082 2084 2087 2095 2097 2098 2099

Meat packing Saus. & prep. meats Poultry & egg proc. Cheese Dry cond. & evap. milk Ice cream & fruit desserts Fluid milk Canned specialties Canned fruit & Veg. Dried fruit & veg. Pickles, sauces, etc. Cereal breakfast foods Rice milling Prep. flour & doughs Wet corn milling Prep. feeds Cookies & crackers Cane sugar Can sugar refining Beet sugar Candy & conf. prods. Chocolate & cocoa Cottonseed oil mill Soybean oil mill Vegetable oil mill An./mar. fats & oils Malt beverages Wines & brandy Food extracts & syrups Roasted coffee Manuf. ice Macaroni & spaghetti Food preparations

2.286∗∗∗ 7.119∗∗∗ 3.864∗∗∗ 1.426∗∗∗ 2.848∗∗∗ 2.637∗∗∗ 9.373∗∗∗ 0.530∗∗∗ 3.613∗∗∗ 1.136∗∗∗ 2.459∗∗∗ 0.531∗∗∗ 7.658 0.550 0.423∗∗∗ 6.043∗∗∗ 0.946∗ 2.063∗∗ 0.184 1.311∗∗∗ 3.254∗∗∗ 0.229 2.246∗∗∗ 0.329 1.653∗∗ 10.551∗∗∗ 0.369∗∗∗ 2.276∗∗∗ 1.276∗∗∗ 0.901 8.587∗∗∗ 3.446∗∗∗ 9.536∗∗∗

0.333 0.815 0.599 0.217 0.556 0.615∗∗∗ 0.790 0.111 0.612 0.301 0.353 0.095 0.587 0.362 0.170 0.899 0.540 0.817 0.254 0.419 0.512 0.140 0.375 0.224 0.635 1.226 0.108 0.239 0.134 0.583 1.240 0.283 1.290

Cost efficiency

s.e.

Output price

−2.244∗∗∗ −0.175 −0.759 0.290 0.804 2.863∗∗ −6.867∗∗∗ 1.317∗∗∗ 5.170∗∗∗ 3.958∗∗∗ −1.383∗∗∗ 1.621∗∗∗ −2.680∗∗ 3.957∗∗∗ 0.448 −0.809 −0.872∗∗∗ −1.721 0.107 −1.423∗ 1.135 1.706∗∗∗ −3.456∗∗∗ −0.062 −2.548∗∗∗ −12.099∗∗∗ 0.719∗∗∗ −0.388 2.193∗∗∗ 0.788 −5.136∗∗∗ −1.862∗∗∗ −3.415∗

0.564 1.294 0.715 0.298 0.863 1.150 1.531 0.193 1.099 0.554 0.239 0.159 1.201 0.773 0.292 1.706 0.881 1.453 0.521 0.834 0.826 0.252 0.565 −0.405 0.635 1.909 0.164 0.333 0.224 1.200 1.731 0.370 1.843

0.042 1.944∗∗∗ 3.105∗∗∗ 1.716∗∗∗ 3.652∗∗∗ 5.500∗∗∗ 2.506∗∗∗ 1.847∗∗∗ 8.783∗∗∗ 5.094∗∗∗ 1.076∗∗∗ 2.152∗∗ −0.022 4.507 0.871∗∗∗ 5.234∗∗∗ 0.074∗∗∗ 0.342 0.291 −0.112 4.389∗∗∗ 1.935∗∗∗ −1.210∗∗∗ 0.267 −0.815∗ −1.548∗∗ 1.083∗∗∗ 1.888∗∗∗ 3.469∗∗∗ 1.689∗∗ 3.451∗∗∗ 1.584∗∗∗ 6.121∗∗∗

s.e. 0.257 0.630 0.354 0.167 0.479 0.643 0.835 0.103 0.588 0.299 0.182 0.084 0.740 0.164 0.864 0.638 0.771 0.273 0.413 0.413 0.125 0.303 0.203 0.354 0.780 0.074 0.189 0.106 0.677 1.185 0.164 1.102

Levels of statistical significance are represented by ∗ (10%), ∗∗ (5%) and ∗∗∗ (1%). The standard errors (s.e.) are indicated next to the estimated coefficients. These results are based on Equation (8).

increases. For the remaining six industries, increases in concentration do not result in significant increases in oligopoly power.8 8 The impacts are particularly large and significant for sausages and prepared meats (2013), poultry and egg processing (2015), fluid milk (2026), candy and confectionary (2064), prepared feeds (2048), animal and marine fats and oils (2077), canned fruits and vegetables (2033), manufactured ice (2097), macaroni and spaghetti (2098), and food preparations (2099).

MARKET POWER AND/OR EFFICIENCY

123

In terms of cost effects, the results indicate that 11 industries (34%) show significant gains (at the 10% level) in cost efficiency with concentration, while eight show a positive and significant increase in cost as concentration rises. The remaining 13 industries do not show a significant impact of concentration on cost efficiency.9 The question then remains how much of these potential cost savings is being passed on and how much is being pocketed by the industries. Combining oligopoly power and cost effects, concentration results in significant decreases in output prices in only three industries (9.4%), with 22 industries (68.7%) showing a net and significant increase in price as concentration rises, while seven (21.9%) show insignificant output price effects at the 10% level. The notable cases where concentration has a benign effect on prices are in the fat and oil industries: 2074, 2076 and 2077. A number of industries show a strong trade-off between oligopoly-power and efficiency effects, where both effects are statistically significant and their magnitudes considerable. In this category fall SICs 2011 (meatpacking), 2022 (cheese), 2026 (fluid milk), 2044 (rice milling), 2074 (soybean oil), 2076 (vegetable oil), 2077 (animal fats and oil), 2097 (manufactured ice), 2098 (macaroni and spaghetti), and food preparations (2099). IV. Concluding Remarks The purpose of this article is two-fold. First, it provides a theory-based empirical model that separates the oligopoly-power and cost-efficiency effects of industrial concentration on output prices. The econometric model adapts the oligopsony model of Azzam (1997) to the oligopoly-power case. A second contribution is that by applying the model to time-series data across U.S. food processing industries, information is provided on the potential impact of further concentration on oligopoly power and cost efficiency within each industry and across industries, which can be useful to policy makers concerned with market structure. From the empirical results, we see some systematic patterns of the impact of industrial concentration in the U.S. food industries. Specifically, we find that further increases in concentration would: (1) Significantly increase oligopoly power; (2) result in cost efficiency in one-third of the industries; and (3) increase output price in nearly every case. These patterns raise some interesting questions. What is and what determines, for example, the critical level of concentration beyond which concentration induces net inefficiency? There are other issues that we do not address here but that are worthy of further attention. Among these are the role of foreign trade and possible simultaneous oligopsony effects on factor prices in certain markets. In spite of the pattern of concentration, there are strong differences among industries. It is timely and rel9 In descending order, the cost efficiency effects that are statistically significant at the 10% level are particularly pronounced in the following industries: animal and marine fats (2077), fluid milk (2026), cottonseed oil (2074), manufactured ice (2097), food preparations (2099), rice milling (2044), vegetable oil (2076), and meat packing (2011).

124

RIGOBERTO A. LOPEZ ET AL.

evant to examine the sources and consequences of these differences more fully. Extending the analysis to incorporate any of these issues or applying it to other industries will provide further insights into the market power and cost-efficiency effects of changes in concentration. Finally, this article shows that NEIO analyses can be made more policy-relevant if they cover market structure and a wider range of industries, as suggested by Bresnahan (1989).

Appendix A: Herfindahl Indexes The Herfindahl–Hirschmann index has only been published for 1982, 1987 and 1992. Additionally, the partial industrial concentration ratios (CR4, CR8, CR20, and CR50) have been published for (then Census years including) 1972 and 1977. Given this paucity of data, H had to be estimated for each food industry for much of the sample period. This process involved two steps: (1) The estimation of the 1972 and 1977 H indexes from partial concentration ratios; (2) interpolation of the H indexes for the inter-Census years. The first step involved the application of entropy to estimate the market shares of the top 50 firms. Using the technique presented by Golan et al. (1996), the market shares were forecasted for all firms encompassed in CR4, CR8, CR20 and CR50. It turns out that the maximum entropy solution for the size distribution of firms yields equal market shares of firms within intervals of these concentration ratios. After forecasting the market share of each firm in 1972 and 1977, we restricted the estimated distribution of market shares to a third-degree polynomial function in order to estimate the most probable market share for each firm. The average Rsquare of the estimated polynomial functions was 95%. From the individual market share, we then estimated the H index as the sum of the squares of market shares resulting from the polynomial distribution. The second step involved regressing the Census-year H indexes on a set of instrumental variables that were available for the entire sample period (Chow and Lin, 1971) such as total number of employees, sales, payroll, and value of shipments per employee for which a complete time-series data were available. These regressions yielded an average R-square of 85%. Then the H indexes were estimated with the predicted values from these regressions. The forecasted values were quite smooth in relation to the H indexes for the Census years. The use of spline functions and exponential smoothing (using the CurveExpert software) did not significantly altered the results. Finally, it should be noted that food-processing industries for which the 1982 Herfindahl indexes were not available (due to reclassification of the SIC codes in 1987) were excluded from the sample. The H indexes used are available from the authors upon request.

MARKET POWER AND/OR EFFICIENCY

125

References Appelbaum, Elie (1982) ‘The Estimation of the Degree of Oligopoly Power’, Journal of Econometrics, 19, 287–299. Azzam, Azzeddine, M. (1997) ‘Measuring Market Power and Cost-Efficiency Effects of Industrial Concentration’, Journal of Industrial Economics, 45, 377–386. Azzam, Azzeddine, and John Schroeter (1995) ‘Tradeoffs between Oligopsony Power and CostEfficiency from Horizontal Consolidation: An Example from Beef Packing’, American Journal of Agricultural Economics, 77, 835–836. Baffes, John, and Utpal Vasavada (1989) ‘On the Choice of Functional Forms in Agricultural Production Analysis’, Applied Economics, 21, 1053–1061. Bartelsman, Eric J., and Wayne Gray (1996) The NBER Manufacturing Productivity Database. National Bureau of Economic Research, Technical Working Paper 205. Bartelsman, Eric J., Randy A. Becker, and Wayne B. Gray (2000) NBER-CES Manufacturing Industry Database (http://www.nber.org/productivity.html). Bhuyan, Sanjib, and Rigoberto A. Lopez (1997) ‘Oligopoly Power in the Food and Tobacco Industries’, American Journal of Agricultural Economics, 7, 1035–1043. Bian, Lin, and D. G. McFetridge (2000) ‘The Efficiencies Defense in Merger Cases: Implications of Alternative Standards’, Canadian Journal of Economics, 33, 297–318. Bresnahan, Timothy F. (1989) ‘Empirical Studies of Industries with Market Power’, in R. Schmalensee and R. Willig, eds., Handbook of Industrial Organization, Vol. 2. Amsterdam: North-Holland. Chow, Gregory C., and An-Noh I. Lin (1971) ‘Best Linear Unbiased Interpolation, Distribution, and Extrapolation of Time Series by Related Series’, Review of Economics and Statistics, 53, 372–375. Cotterill, Ronald W. (1986) ‘Market Power in the Retail Food Industry: Evidence from Vermont’, Review of Economics and Statistics, 68, 379–386. Demsetz, Harold (1973) ‘Industry Structure, Market Rivalry, and Public Policy’, Journal of Law and Economics, 16, 1–9. Dickson, Vaughn (1994) ‘Aggregate Industry Cost Functions and the Herfindahl Index’, Southern Journal of Economics, 61, 445–452. Gisser, Micha (1982) ‘Welfare Implications of Oligopoly in U.S. Food Manufacturing’, American Journal of Agricultural Economics, 63, 616–624. Golan, A., George Judge, and Jeffrey M. Perloff (1996) ‘Estimating the Size Distribution of Firms Using Government Summary Statistics’, Journal of Industrial Economics, 44, 69–80. Martin, Steven (1988) ‘Market Power and/or Efficiency?’, Review of Economics and Statistics, 60, 331–335. Morrison-Paul, Catherine J. (2000) ’Cost Economies and Market Power: The Case of the U.S. Meat Packing Industry’, Review of Economics and Statistics, 82, in press. Nevo, Aviv (2000) ‘Measuring Market Power in the Ready-to-Eat Cereal Industry’, Econometrica, 68, forthcoming. Olson, D. O., and Y. N. Shieh (1989) ‘Estimating Functional Forms in Cost Prices’, European Economic Review, 33, 1445–1461. Peltzman, Samuel (1977) ‘The Gains and Losses from Industrial Concentration’, Journal of Law and Economics, 20, 229–263. Perry, Martin K. (1984) ‘Scale Economies, Imperfect Competition, and Public Policy’, Journal of Industrial Economics, 32, 313–333. Rosenbaum, David I. (1994) ‘Efficiency v. Collusion: Evidence Cast in Cement’, Review of Industrial Organization, 9, 379–392. Salinger, Michael A. (1990) ‘The Concentration-Margin Relationship Revisited’, Brookings Papers on Economic Activity: Microeconomics. Washington, D.C.: Brookings Institution.

126

RIGOBERTO A. LOPEZ ET AL.

Schroeter, John R. (1988) ‘Estimating the Degree of Market Power in the Beef Packing Industry’, Review of Economics and Statistics, 70, 158–162. Schroeter, John R., and Azzeddine M. Azzam (1990) ‘Measuring Market Power in Multi-Product Oligopolies: The U.S. Meat Industry’, Applied Economics, 22, 1365–1376. Stalhammar, N. O. (1991) ‘Domestic Market Power and Foreign Trade’, International Journal of Industrial Organization, 9, 407–424. Stigler, George. (1964) ‘A Theory of Oligopoly’, Journal of Political Economy, 72, 44–61. U.S. Department of Labor (1997) BLS Handbook of Methods, Bureau of Labor Statistics, Bulletin 2490. Williamson, O. (1968) ‘Economies as an Antitrust Defense: The Welfare Trade-Offs’, American Economic Review, 58, 18–36.